Design of an Automatic Subtype Classifier of Breast Carcinoma Using XResNet with Cyclic Scheduling
Vivek Harshey, Amar Partap Singh Pharwaha · 2023
This paper reports the design of an automatic system to classify breast cancer subtypes using an XResNet deep learning model. The prior solutions to this problem suffer from low accuracy or cost of training. Our designed system effectively classifies the breast cancer histopathology patches cropped from whole slide images while being affordable and accurate. The reported system effectively detects cancer's localized or invasive nature. The patient's prognosis and treatment schedule depend on the multi-classification of the images. The presented system achieves state-of-the-art accuracy on the multiclass BACH dataset. On the BACH dataset, we achieved the highest accuracy of 89.95% and balanced accuracy of 87.8% in only 30 epochs matching the state-of-the-art systems. Comparisons of the designed system with recent works establish the model's efficacy, making it worthy for the state-of-the-art breast cancer diagnosis routine. The designed system can be a valuable tool for medical professional for improving the survival of patients.